init

extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/exec/ControllerImpl.java

@@ -2005,7 +2005,8 @@ private static Pair<List<DimensionSchema>, List<AggregatorFactory>> makeDimensio
                 DimensionSchemaUtils.createDimensionSchema(
                     outputColumnName,
                     type,
-                    MultiStageQueryContext.useAutoColumnSchemas(query.context())
+                    MultiStageQueryContext.useAutoColumnSchemas(query.context()),
+                    MultiStageQueryContext.isIngestStringArraysAsMVDs(query.context())
                 )
             );
           } else if (!isRollupQuery) {
@@ -2054,7 +2055,8 @@ private static void populateDimensionsAndAggregators(
           DimensionSchemaUtils.createDimensionSchema(
               outputColumn,
               type,
-              MultiStageQueryContext.useAutoColumnSchemas(context)
+              MultiStageQueryContext.useAutoColumnSchemas(context),
+              MultiStageQueryContext.isIngestStringArraysAsMVDs(context)
           )
       );
     }

extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/input/external/ExternalInputSliceReader.java

@@ -121,6 +121,7 @@ private static Iterator<SegmentWithDescriptor> inputSourceSegmentIterator(
                     DimensionSchemaUtils.createDimensionSchema(
                         column,
                         signature.getColumnType(column).orElse(null),
+                        false,
                         false
                     )
             ).collect(Collectors.toList())

extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/DimensionSchemaUtils.java

@@ -43,12 +43,13 @@ public class DimensionSchemaUtils
   public static DimensionSchema createDimensionSchema(
       final String column,
       @Nullable final ColumnType type,
-      boolean useAutoType
+      boolean useAutoType,
+      boolean writeStringArraysAsMVDs
   )
   {
     if (useAutoType) {
       // for complex types that are not COMPLEX<json>, we still want to use the handler since 'auto' typing
-      // only works for the 'standard' built-in typesg
+      // only works for the 'standard' built-in types
       if (type != null && type.is(ValueType.COMPLEX) && !ColumnType.NESTED_DATA.equals(type)) {
         final ColumnCapabilities capabilities = ColumnCapabilitiesImpl.createDefault().setType(type);
         return DimensionHandlerUtils.getHandlerFromCapabilities(column, capabilities, null)
@@ -57,7 +58,7 @@ public static DimensionSchema createDimensionSchema(
 
       return new AutoTypeColumnSchema(column);
     } else {
-      // if schema information not available, create a string dimension
+      // if schema information is not available, create a string dimension
       if (type == null) {
         return new StringDimensionSchema(column);
       }
@@ -74,7 +75,11 @@ public static DimensionSchema createDimensionSchema(
         case ARRAY:
           switch (type.getElementType().getType()) {
             case STRING:
-              return new StringDimensionSchema(column, DimensionSchema.MultiValueHandling.ARRAY, null);
+              if (writeStringArraysAsMVDs) {
+                return new StringDimensionSchema(column, DimensionSchema.MultiValueHandling.ARRAY, null);
+              } else {
+                return new AutoTypeColumnSchema(column);
+              }
             case LONG:
             case FLOAT:
             case DOUBLE:

extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/util/MultiStageQueryContext.java

@@ -73,6 +73,10 @@
  * {@link org.apache.druid.segment.AutoTypeColumnSchema} for all 'standard' type columns during segment generation,
  * see {@link DimensionSchemaUtils#createDimensionSchema} for more details.
  *
+ * <li><b>ingestStringArraysAsMVDs</b>: Flag to ingest the string arrays using string dimension handlers, which generates
+ * MVDs instead of {@code ARRAY<STRING>}. The flag is undocumented, and provided to preserve legacy behaviour.
+ * see {@link DimensionSchemaUtils#createDimensionSchema} for more details.
+ *
  * </ol>
  **/
 public class MultiStageQueryContext
@@ -121,6 +125,11 @@ public class MultiStageQueryContext
   public static final String CTX_INDEX_SPEC = "indexSpec";
 
   public static final String CTX_USE_AUTO_SCHEMAS = "useAutoColumnSchemas";
+  public static final boolean DEFAULT_USE_AUTO_SCHEMAS = false;
+
+  public static final String CTX_INGEST_STRING_ARRAYS_AS_MVDS = "ingestStringArraysAsMVDs";
+  public static final boolean DEFAULT_INGEST_STRING_ARRAYS_AS_MVDS = false;
+
 
   private static final Pattern LOOKS_LIKE_JSON_ARRAY = Pattern.compile("^\\s*\\[.*", Pattern.DOTALL);
 
@@ -243,7 +252,12 @@ public static IndexSpec getIndexSpec(final QueryContext queryContext, final Obje
 
   public static boolean useAutoColumnSchemas(final QueryContext queryContext)
   {
-    return queryContext.getBoolean(CTX_USE_AUTO_SCHEMAS, false);
+    return queryContext.getBoolean(CTX_USE_AUTO_SCHEMAS, DEFAULT_USE_AUTO_SCHEMAS);
+  }
+
+  public static boolean isIngestStringArraysAsMVDs(final QueryContext queryContext)
+  {
+    return queryContext.getBoolean(CTX_INGEST_STRING_ARRAYS_AS_MVDS, DEFAULT_INGEST_STRING_ARRAYS_AS_MVDS);
   }
 
   /**